blob: 75fedd823db6d65245cf90d29f91748facdb583e (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
|
import os
from loguru import logger as log
from sentence_transformers import CrossEncoder
from rag.generator.prompt import Prompt
from rag.retriever.rerank.abstract import AbstractReranker
class Reranker(metaclass=AbstractReranker):
def __init__(self) -> None:
self.model = CrossEncoder(os.environ["RERANK_MODEL"])
self.top_k = int(os.environ["RERANK_TOP_K"])
self.relevance_threshold = float(os.environ["RETRIEVER_RELEVANCE_THRESHOLD"])
def rank(self, prompt: Prompt) -> Prompt:
if prompt.documents:
results = self.model.rank(
query=prompt.query,
documents=[d.text for d in prompt.documents],
return_documents=False,
top_k=self.top_k,
)
ranking = list(
filter(
lambda x: x.get("score", 0.0) > self.relevance_threshold, results
)
)
log.debug(
f"Reranking gave {len(ranking)} relevant documents of {len(prompt.documents)}"
)
prompt.documents = [
prompt.documents[r.get("corpus_id", 0)] for r in ranking
]
return prompt
|
